Forecasting commodity futures using Principal Component Analysis and Copula

Abstract: The ever ongoing battle to beat the market is in this thesis fought with the help of mathematics with a way to reduce the information to its core. It is called PCA, Principal Component Analysis. This is used to build a model of future commodity prices. To assist PCA, Copula is used - a sort of mathematical glue which can bring multiple distributions together and represented as one. The data used is 5 years of prices for Brent Oil, WTI Oil, Gold, Copper and Aluminium. The model parameters are tted to 2.5 years of data and then tested on the remaining 2.5 years. MLE, Maximum Likelihood Estimation, was used for parameter estimation and distributions that were found tting were logistic and Student's T distribution Cramer-von Mises tests were used to determine that T Copula was the most suitable Copula. The main results are that the mathematical estimations t well and prot can be generated, but with a low Sharpe Ratio.